Slides for Chapter 4: Interprocess Communication

Transcription

1 Slides for Chapter 4: Interprocess Communication From Coulouris, Dollimore, Kindberg and Blair Distributed Systems: Concepts and Design Edition 5, Addison-Wesley 2012 Text extensions to slides David E. Bakken, 2012

2 Next 2 Chapters: Communication Aspects of MW Chap 4: IPC API for Internet protocols External data representation and marshalling Multicast communications Network virtualization: overlay networks Case Study: MPI Chapter 5: Remove invocation paradigms Request-reply protocols Remote procedure call Remove method invocation Case study: Java RMI 2

3 API for Internet Protocols [4.2] 4.2.1: How to implement send/receive of Sec : Sockets 4.2.3: UDP in Java 4.2.4: TCP in Java 3

4 Figure 4.1 Middleware layers APIs for Internet protocols UDP: message passing abstraction (incl. multicast) TCP: stream processing Text Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

5 Characteristics of IPC [4.2.1] Message passing supported by send+receive to endpoints ( message destination in text) Synchronous and asynchronous communication Associate a queue with each endpoint Senders add msg to remote queue; Receivers remove msg from local queue Send and receive both have synch. (blocking) and asynch. (non-blocking) Non-blocking receive not always supported cause more complex for app Can combine in nice ways SR Language from U. Arizona 1990s, Sender Receiver Blocking wait Proc creation Asynch Asynch. Msg passing Dynamic proc creation Synch Rendezvous RPC 5

6 Characteristics of IPC (cont.) Message destinations Endpoint on internet is (IP address, (remote) port #) Local port created for both sides to access One receiving process per port Process can use multiple ports to receive messages Transparency support: look up endpoint from name (Sec 5.4.2) Reliability Reliable: msg delivered despite reasonable packets lost Integrity support: msgs must be uncorrupted and no dups More above too to verify sender etc Ordering Some apps require sender order (FIFO) Layer above UDP (if multicast), or with TCP 6

7 Figure 4.2 Sockets and ports socket any port agreed port socket client message other ports server Internet address = Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012 Internet address = Socket abstraction: hide steps in lower-level protocols etc To receive: socket bound to local port Can use same socket to both send and receive large number (!!) of ports, 2 16 Anyone can send to a socket if know endpoint Can not have multiple receivers (except for mulitcast..)

8 Java API for IP Addresses Class InetAddress (both IPv4 and IPv6) InetAddress acomputer = InetAddress.getByName( foo.bar.edu ) Can throw UnknownHostException 8

9 UDP datagram communication [4.2.3] Sent without ACKs or retries Issues Message size: receiver gives fixed-sized buffer, if msg big trunc. Blocking: receive blocks (unless timeout set), send rtns quickly Receiver can get msg later if not blocked (queued) Timeouts: can set, choosing really hard Receive from any: no origin (sender) specified, endpoint in header Failure model: omissions, ordering Uses of UDP DNS VOIP and other video/audio Higher-level multicast with stronger properties 9

10 Java API for UDP datagrams DatagramPacket class Sending constructor takes array of bytes, length, endpoint Another constructor for receiving msgs: array of bytes, length DatagramPacket.getData: receiver gets buffer DatagramSocket class Constructor: port (also no-port: choose any free local port) Send() and receive() setsotimeout: for receive, if times out throws InterruptedIOException Connect: connect to remote endpoint 10

11 Figure 4.3 UDP client sends a message to the server and gets a reply import java.net.*; import java.io.*; public class UDPClient{ public static void main(string args[]){ // args give message contents and server hostname DatagramSocket asocket = null; try { asocket = new DatagramSocket(); byte [] m = args[0].getbytes(); InetAddress ahost = InetAddress.getByName(args[1]); int serverport = 6789; DatagramPacket request = new DatagramPacket(m, m.length(), ahost, serverport); asocket.send(request); // send message to the remote endpoint byte[] buffer = new byte[1000]; DatagramPacket reply = new DatagramPacket(buffer, buffer.length); asocket.receive(reply); // blocking wait for reply System.out.println("Reply: " + new String(reply.getData())); }catch (SocketException e){system.out.println("socket: " + e.getmessage()); }catch (IOException e){system.out.println("io: " + e.getmessage());} }finally {if(asocket!= null) asocket.close();} } } Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

12 Figure 4.4 UDP server repeatedly receives a request and sends it back to the client import java.net.*; import java.io.*; public class UDPServer{ public static void main(string args[]){ DatagramSocket asocket = null; try{ asocket = new DatagramSocket(6789); byte[] buffer = new byte[1000]; while(true){ // typical infinite server waiting loop: get request, send reply DatagramPacket request = new DatagramPacket(buffer, buffer.length); asocket.receive(request); DatagramPacket reply = new DatagramPacket(request.getData(), request.getlength(), request.getaddress(), request.getport()); asocket.send(reply); } }catch (SocketException e){system.out.println("socket: " + e.getmessage()); }catch (IOException e) {System.out.println("IO: " + e.getmessage());} }finally {if(asocket!= null) asocket.close();} } } Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

13 TCP stream communication [4.2.4] Hides many details beyhind socket abstraction Message sizes: sender chooses how much data to read or write Underlying impl decides when to send packets Can flush/synch to force a send (why need?) Lost packets Flow control Packet duplication and ordering Message destinations (once socket set up) Server creates listening socket Client calls connect( ) Server calls accept( ) 13

14 TCP stream communication (cont) API Assumes for setup one side client other server After that bidirectional with no distinction (both input and output stream in socket) Listening socket maintains queue of connect requests Issues with TCP (and stream communication) Matching data items: need to agree on data types (UDP too) Blocking: like UDP Threads: server usually forks new process for each client (why?) 14

15 TCP stream communication (cont) Failure model Checksums detects and rejects corrupted packets Sequence numbers: detect and reject duplicate packets If too many lost packets, socket declared to be closed I.e., not (very) reliable communication Uses of TCP (lots): HTTP, FTP, Telnet, SMTP 15

16 Java API for TCP streams ServerSocket class: for listening to connect requests Method accept gets connect request or blocks if none queued Returns a Socket object to communication with the client Socket: for pair to communicate with Client uses constructor to create a given endpoint Can throw UnknownHostexception or IOException getinputstream and getoutputstream to access streams 16

17 Figure 4.5: TCP client makes connection to server, sends request and receives reply import java.net.*; import java.io.*; public class TCPClient { public static void main (String args[]) { // arguments supply message and hostname of destination Socket s = null; try{ int serverport = 7896; s = new Socket(args[1], serverport); DataInputStream in = new DataInputStream( s.getinputstream()); DataOutputStream out = new DataOutputStream( s.getoutputstream()); out.writeutf(args[0]); // UTF is a string encoding see Sn 4.3 String data = in.readutf(); System.out.println("Received: "+ data) ; }catch (UnknownHostException e){ System.out.println("Sock:"+e.getMessage()); }catch (EOFException e){system.out.println("eof:"+e.getmessage()); }catch (IOException e){system.out.println("io:"+e.getmessage());} }finally {if(s!=null) try {s.close();}catch (IOException e){system.out.println("close:"+e.getmessage());}} } } Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

18 Figure 4.6 TCP server makes a connection for each client and then echoes the client s request import java.net.*; import java.io.*; public class TCPServer { public static void main (String args[]) { try{ int serverport = 7896; ServerSocket listensocket = new ServerSocket(serverPort); while(true) { Socket clientsocket = listensocket.accept(); Connection c = new Connection(clientSocket); } } catch(ioexception e) {System.out.println("Listen :"+e.getmessage());} } } // this figure continues on the next slide Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

19 Figure 4.6 continued class Connection extends Thread { DataInputStream in; DataOutputStream out; Socket clientsocket; public Connection (Socket aclientsocket) { try { clientsocket = aclientsocket; in = new DataInputStream( clientsocket.getinputstream()); out =new DataOutputStream( clientsocket.getoutputstream()); this.start(); } catch(ioexception e) {System.out.println("Connection:"+e.getMessage());} } public void run(){ try { // an echo server String data = in.readutf(); out.writeutf(data); } catch(eofexception e) {System.out.println("EOF:"+e.getMessage()); } catch(ioexception e) {System.out.println("IO:"+e.getMessage());} } finally{ try {clientsocket.close();}catch (IOException e){/*close failed*/}} } } Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

20 External data representation and marshalling [4.3] Procedures/methods called with params, msgs take buffers Marshalling does this translation, unmarshalling reverses External data representations describe how Endian-ness, ASCI or Unicode text, etc Two main techniques Neutral format Sender s format ( receiver makes right ) 20

21 Approaches to marshalling and external data representation 1. CORBA s common data rep. (CDR): structs, primitives 2. Java serialization: object or tree of objects 3. XML: textual description of data Comparisons CDR, Java: middleware layer, XML more for hand coding in app CDR, Java: binary form, XML text CDR: only values (sort of), Java, XML: type info XML larger, more error prone than automatic marshalling by middleware compiler Other possibilities (more lightweight ) Google protocol buffers (20.4.1): describe stored&transmitted data JavaScript Object Notation (JSON) 21

22 CORBA s Common Data Representation (CDR) [4.3.1] All 15 primitive types: short, long, boolean, any Defn s for both big- and little-endian (sent in sender s order; tag) Other types straightforward: IEEE floats, chars agreed between client and server Constructed/composite types (Fig 4.7, next) Primitive types that make them up added in a byte sequence in a given order Marshalling generated automatically from IDL 22

23 Figure 4.7 CORBA CDR for constructed types Type sequence string array struct enumerated union Representation length (unsigned long) followed by elements in order length (unsigned long) followed by characters in order (can also can have wide characters) array elements in order (no length specified because it is fixed) in the order of declaration of the components unsigned long (the values are specified by the order declared) type tag followed by the selected member Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

24 Figure 4.8 CORBA CDR message index in sequence of bytes "Smit" "h " 6 "Lond" "on " bytes notes on representation length of string Smith length of string London unsigned long struct Person { string name; string place; unsigned long year; }; The flattened form represents a Person struct with value: { Smith, London, 1984} Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

25 Java Object Serialization [4.3.2] Java class equivalent to CORBA Person struct: public class Person implements Serializable { private String name; private String place; private int year; public Person(String aname, String aplace, int ayear) { name = aname; place = aplace; year = ayear; } // followed by methods for accessing the instance vars }// class Person 25

26 Java Serialization (cont.) Handles: serialized references to other objects Reflection:used by serialization to find class name of object to be serialized, when deserialized to create class (Read rest of details in text, not covering) 26

27 Figure 4.9 Indication of Java serialized form Serialized values Explanation Person 8-byte version number h0 class name, version number 3 int year java.lang.string name: java.lang.string place: number, type and name of instance variables Smith 6 London h1 values of instance variables The true serialized form contains additional type markers; h0 and h1 are handles Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

28 Extensible Markup Language (XML) [4.3.3] Markup language: encodes both text and its structure HTML: for web pages XML: for structured documents for web Both: derived from (very complext) SGML XML namespaces: provided for scoping names (avoid name collisions) XML schemas: define elements and attributes for a doc, nesting, order and number of elements, etc Overview in this lecture, read details in text Gory details not testable (don t memorize minutia) But should really have intuition into difference from CORBA and Java: purpose, why design decisions made, efficiency, readability, other comparisons 28

29 Figure 4.10 XML definition of the Person structure <person id=" "> <name>smith</name> <place>london</place> <year>1984</year> <!-- a comment --> </person > Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

30 Figure 4.11 Illustration of the use of a namespace in the Person structure <person pers:id=" " xmlns:pers = " <pers:name> Smith </pers:name> <pers:place> London </pers:place > <pers:year> 1984 </pers:year> </person> Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

31 Figure 4.12 An XML schema for the Person structure <xsd:schema xmlns:xsd = URL of XML schema definitions > <xsd:element name= "person" type ="persontype" /> <xsd:complextype name="persontype"> <xsd:sequence> <xsd:element name = "name" type="xs:string"/> <xsd:element name = "place" type="xs:string"/> <xsd:element name = "year" type="xs:positiveinteger"/> </xsd:sequence> <xsd:attribute name= "id" type = "xs:positiveinteger"/> </xsd:complextype> </xsd:schema> Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

32 Remote object references [4.3.4] (Only applies to CORBA &Java: distributed object model) NOT XML Remove object reference: identifier valid thru a DS Generated so unique over space and time Lots of processes in a DS! Must not reuse 32

33 Figure 4.13 Representation of a remote object reference 32 bits 32 bits 32 bits 32 bits Internet address port number time object number interface of remote object Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

34 CORBA Object References (not in texbook) Object reference Opaque handle for client to use Identifies exactly one CORBA object IOR == Interoperable Object Reference References may be passed among processes on different hosts As parameters, return values, or stringified ORB will marshall ORB on receiver side unmarshalling will create a proxy return a pointer to it Basically functions as a remote pointer that works across heterogeneity in language, OS, net, vendor, 34

35 CORBA Object References (cont.) IOR: Repository ID Profile for Protocol1 Profile for Protocol2 IIOP Profile: TAG_INTERNET_IOP IIOP version Host addr. Port Object Key Components Object Key (one possible implementation): POA ID Object ID Object Key Opaque to client ORB-specific Object ID Can be created by user or POA Components: Optional data e.g., alternate endpoint, info for security policies, etc 35

36 Multicast Communication [4.4] Point-to-point communications not great for process groups Multicast: 1:many communications, many uses Fault-tolerance based on replicated services Discovering services in spontaneous networking Better performance through replicated data (multicast updates) Propagation of event notices (Facebook, implement pub-sub) 36

37 IP Multicast (IPMC) [4.4.1] One implementation of multicast, using UDP not TCP Use normal sockets to join (receiving) group Multicast routers: can send to multiple LANs (use its mcast) Multicast addresses Permanent: assigned by IANA, exist even if no members Temporary: come and go dynamically Failure model: Same as UDP (omission) Some group members may receive, others not AKA unreliable multicast (reliable multicast in Chapter 15, for 562) 37

38 Java API for IP Multicast Class MulticastSocket Subclass of DatagramSocket joingroup( ) leavegroup( ) 38

39 Figure 4.14: Multicast peer joins a group and sends and receives datagrams import java.net.*; import java.io.*; public class MulticastPeer{ public static void main(string args[]){ // args give message contents & destination multicast group (e.g. " ") MulticastSocket s =null; try { InetAddress group = InetAddress.getByName(args[1]); s = new MulticastSocket(6789); s.joingroup(group); byte [] m = args[0].getbytes(); DatagramPacket messageout = new DatagramPacket(m, m.length, group, 6789); s.send(messageout); // this figure continued on the next slide Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

40 Figure 4.14 continued } } // get 3 messages from others in group byte[] buffer = new byte[1000]; for(int i=0; i< 3; i++) { DatagramPacket messagein = new DatagramPacket(buffer, buffer.length); s.receive(messagein); System.out.println("Received:" + new String(messageIn.getData())); } s.leavegroup(group); }catch (SocketException e){system.out.println("socket: " + e.getmessage()); }catch (IOException e){system.out.println("io: " + e.getmessage());} }finally {if(s!= null) s.close();} Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education 2012

41 Reliability and ordering of multicast [4.4.2] IPMC: dropped msgs, partial delivery to group, no ordering Effects on different apps? Fault-tolerance based on replicated services E.g., keep replicas with same state, multicast requests What happens with failures above? Discovering services in spontaneous networking What happens with failures above? Better performance through replicated data (multicast updates) What happens with failures above? Propagation of event notices (Facebook, implement pub-sub) What happens with failures above? 41

42 Network virtualization: overlay networks [4.4] Some applications need (much) more advanced delivery services than Internet protocols provide End-to-end argument says to not push functions down here Network virtualization: construct many different virtual networks over Internet Support specific services needed Answers end-to-end argument: app-specific virtual network 42

43 Overlay networks [4.5.1] Overlay network: virtual network consisting of topology of virtual nodes and virtual links (above underlay network s) Tailor services to specific app (e.g., multimedia content) More efficient in some network environments (ad hoc) Add more features: multicast, secure communications, Can redefine addressing, protocols, routing Advantages Add services without having to change (and standardize) underlay Encourage experimentation Can exist with other overlays (same kind or different) 43

44 Overlay networks (cont) Disadvantages: extra level of indirection Placement (overlay underlay) is key for efficiency add to complexity Examples in book Skype next Chap 10: P2P file sharing, distributed hash tables Chap 19: mobile/ubiquitous: ad hoc and disruption-tolerant Chap 20: multimedia streaming 44

45 Figure 4.15 Types of overlay table continues on the next slide Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education

46 Figure 4.15 (continued) Types of overlay Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education

47 Skype [4.5.2] P2P VOIP overlay network (instant msgs, video, telephony) Addresses: skype username or phone number Architecture: P2P Ordinary user machines hosts Well enabled/connected hosts: super node Authenticate users over well-known login server, gives them super node Supernodes goal: search for users efficiently Direct voice connection between two parties Signalling: TCP Call: UDP or TCP (latter only to circumvent firewalls) Codecs key: optimized for 32 kbps 47

48 Figure 4.16 Skype overlay architecture Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concepts and Design Edn. 5 Pearson Education

49 Case Study: MPI [4.6] Started in grid computing (and influenced it) NOT Covering: required for 564 (not for 464) 49